[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [cobalt-users] Load Average High
- Subject: Re: [cobalt-users] Load Average High
- From: Jason Woods <cobalt-ml@xxxxxxxxxxxxxxx>
- Date: Fri Mar 22 09:49:01 2002
- Organization: Oakland Corporation
- List-id: Mailing list for users to share thoughts on Cobalt products. <cobalt-users.list.cobalt.com>
baltimoremd@xxxxxxxxxxxxxxx wrote:
>
> On Thu, 21 Mar 2002, E.B. Dreger wrote:
>
> > > Date: Thu, 21 Mar 2002 01:06:22 -0500 (EST)
> > > From: baltimoremd@xxxxxxxxxxxxxxx
> >
<snip>
>
> > Can you provide more details?
>
> I think it's a customer's cgi script that's being a resource pig, since I
> see some lines in the kernel log ala
> Mar 21 00:08:48 ns1 kernel: VM: do_try_to_free_pages failed for
> prebot.cgi...
> Mar 21 00:08:48 ns1 kernel: VM: do_try_to_free_pages failed for
> prebot.cgi...
> Mar 21 00:08:48 ns1 kernel: VM: do_try_to_free_pages failed for httpd...
> Mar 21 00:08:48 ns1 kernel: VM: do_try_to_free_pages failed for
> prebot.cgi...
>
> So, I'm half tempted to politely ask the guy to move on...since I'm not
> sure he's savy enough to make the script less of a burden.
>
> Thom
Sounds like you ran out of memory. His script might have just happened
to be called during the swap-a-thon that was happening on the server.
The fastest way to kill a server is to eat up enough memory to start
swapping. Run whatever page calls your customer's script and watch
top. See how much memory it is using, and how many processes it
spawns. It it only uses about 10 MB (normal for somewhat complicated
perl scripts on my server), and only has 1 or 2 processes going, it
isn't a big deal, and probably was an innocent by-stander at the time in
question. I would say that a combination of majordomo dueling out a
large list, sendmail taking care of it, apache running, and the cgi
script all caused this.
One thing I am not sure about, and I wouldn't be able to see without a
ps output, is why there were 130 processes running. I am assuming
majordomo was at fault here. Try sending mail out to that mailing list
while watching top and see what happens. I would guess that your total
process number would jump dramatically.
How much RAM do you have in your RaQ? I would definitely suggest at
-least- 256 MB.
--
Jason Woods
Oakland Corporation
jwoods@xxxxxxxxxxxxxxx